基于高光谱成像技术的牛肉大理石花纹的评估  

Analysis of Beef-marbling Grade Using Hyperspectral Imaging Technology

作　　者：高晓东[1] 吴建虎[1] 彭彦昆[1] 陈菁菁[1] 陶斐斐[1] 

机构地区：[1]中国农业大学工学院,北京100083

出　　处：《农产品加工（下）》2009年第10期33-37,共5页Farm Products Processing

基　　金：国家自然科学基金资助项目(30771244)

摘　　要：利用高光谱扫描成像技术评估牛肉大理石花纹。组建了高光谱线扫描成像系统,采集牛肉样品在400-1100nm波段的高光谱反射图像。通过牛肉脂肪和瘦肉在各个波段处反射值比的最大值,确定530nm为特征波段。提取特征波段处大理石花纹的3个特征参数（大颗粒脂肪密度、中等颗粒脂肪密度和小颗粒脂肪密度）,使用特征参数分别建立多元线性回归模型（MLR）和正则判定函数模型,对大理石花纹分级和等级预测,用全交叉验证方法验证模型的准确性。MLR模型对大理石花纹等级的预测决定系数R2=0.92,预测标准差为SECV=0.45;总的分级准确率是84.8%;正则判定函数对大理石花纹等级判定准确性较低,为78.8%。研究表明,将高光谱成像技术应用于牛肉大理石花纹等级评定是可行的。The objective of this study was to analyze the beef-marbling grade using hyperspectral imaging technology. Hyperspectral scanning imaging system was developed to collect hyperspectral images in the spectral region of 400-1 100 nm. The max ratio of gray value of fat and lean in each bands was used to select the characteristic bands, images from 530 nm was used to segment beef-marbling. Three characteristic parameters of beef-marbling such as big fat area, medium fat area and small fat area were extracted. Then the three parameters were used to establish prediction model by multiple linear regression and canonical discriminant function methods. The MLR model result shows R^2=0.92, SECV=0.45, the accuracy of classification is 84.8% ; and canonical discriminant functions showed classification accuracy of 78.8%. This research demonstrated that hyperspectral imaging technology is useful for nondestructive determination of beef tenderness.

关 键 词：牛肉大理石花纹 高光谱成像 图像处理 

分 类 号：S123[农业科学—农业基础科学] TS207[轻工技术与工程—食品科学]